Gemini 3 Flash: frontier intelligence built for speed

Google DeepMind · 7 months ago

Google released Gemini 3 Flash, a faster and cheaper version of its Gemini 3 model, making it the default model across the Gemini app and AI Mode in Search. The model processes over 1 trillion tokens per day on Google's API, scores 90.4% on GPQA Diamond reasoning benchmarks, runs 3 times faster than its predecessor while costing $0.50 per million input tokens, and uses 30% fewer tokens on average than the prior model. Users globally now access Gemini 3's reasoning capabilities for free, while developers gain a tool for building agentic workflows and interactive applications with high-frequency responsiveness.